Post on 21-Dec-2015
Signals: Implications for Business Cycles and
Monetary Policy
Lawrence Christiano,Cosmin Ilut,
Roberto Motto, and Massimo Rostagno
Objective• Estimate a model in which technology shocks
are partially anticipated– ‘Normal’ technology shock:
– Shock considered here (J Davis):
• Evaluate importance of for business cycles
• Explore implications of for monetary policy.
at aat 1 t
at aat 1
‘recent information’
t t 11 t 2
2 t 33 t 4
4
‘earlier information’
t 55 t 6
6 t 77 t 8
8
t ii
t ii
Outline• Estimation
– Results– ‘Excessive optimism’ and 2000 recession
• Implications for monetary policy– Monetary policy causes economy to over-
react to signals....inadvertently creates ‘boom-bust’
Model• Features (version of CEE)
– Habit persistence in preferences
– Investment adjustment costs in change of investment
– Variable capital utilization
– Calvo sticky (EHL) wages and prices
• Non-optimizers:
• Probability of not adjusting prices/wages:
Pit Pi,t 1, Wj,t zWj,t 1
p, w
Observables and Shocks• Six observables:
– output growth, – inflation, – hours worked, – investment growth, – consumption growth, – T-bill rate.
• Sample Period: 1984Q1 to 2007Q1
E tj l 0
1
1.03 1/4
lpreference shock
c,t l logCt l bCt l 1 L
l t l,j2
2
Kt 1 1 0.02Kt 1 S
marginal (in-) efficiency of investment I,t I t
I t 1I t
Yt 0
1Yjt
1 f,t dj
markup shock f,t
, Yj,t zt exptechnology shock
at L j,t
1
utKj,t , zt exp zt
log R tR
log R t 1
R 1 1
Ra log t 1
ay4
logyty tM
E tj l 0
1
1.03 1/4
lpreference shock
c,t l logCt l bCt l 1 L
l t l,j2
2
Kt 1 1 0.02Kt 1 S
marginal (in-) efficiency of investment I,t I t
I t 1I t
Yt 0
1Yjt
1 f,t dj
markup shock f,t
, Yj,t zt exptechnology shock
at L j,t
1
utKj,t , zt exp zt
log R tR
log R t 1
R 1 1
Ra log t 1
ay4
logyty tM
E tj l 0
1
1.03 1/4
lpreference shock
c,t l logCt l bCt l 1 L
l t l,j2
2
Kt 1 1 0.02Kt 1 S
marginal (in-) efficiency of investment I,t I t
I t 1I t
Yt 0
1Yjt
1 f,t dj
markup shock f,t
, Yj,t zt exptechnology shock
at L j,t
1
utKj,t , zt exp zt
log R tR
log R t 1
R 1 1
Ra log t 1
ay4
logyty tM
E tj l 0
1
1.03 1/4
lpreference shock
c,t l logCt l bCt l 1 L
l t l,j2
2
Kt 1 1 0.02Kt 1 S
marginal (in-) efficiency of investment I,t I t
I t 1I t
Yt 0
1Yjt
1 f,t dj
markup shock f,t
, Yj,t zt exptechnology shock
at L j,t
1
utKj,t , zt exp zt
log R tR
log R t 1
R 1 1
Ra log t 1
ay4
logyty tM
E tj l 0
1
1.03 1/4
lpreference shock
c,t l logCt l bCt l 1 L
l t l,j2
2
Kt 1 1 0.02Kt 1 S
marginal (in-) efficiency of investment I,t I t
I t 1I t
Yt 0
1Yjt
1 f,t dj
markup shock f,t
, Yj,t zt exptechnology shock
at L j,t
1
utKj,t , zt exp zt
log R tR
log R t 1
R 1 1
Ra log t 1
ay4
logyty tM
Shock representationsmarkup
log f,t f
f log f,t 1
f f ,t
discount rate
log c,t c log c,t 1 c,t
efficiency of investment
log I,t I log I,t 1 I,t
technology
at aat 1 iid
t
iid t 1
1
iid t 2
2
iid t 3
3
iid t 4
4
iid t 5
5
iid t 6
6
iid t 7
7
iid t 8
8
monetary policy
tM M t 1M u,t.
Variance Decomposition, Technology Shocks
variable t i 18 t ii t t
i 14 t ii
i 58 t ii
consumption growth 46.6 7.0 24.1 22.5
investment growth 16.1 2.3 8.2 7.9
output growth 45.4 6.2 23.1 22.3
log hours 45.3 5.5 20.0 25.3
inflation 49.0 7.0 23.8 25.2
interest rate 52.1 7.1 24.9 27.2
bi
Big!
Variance Decomposition, Technology Shocks
variable t i 18 t ii t t
i 14 t ii
i 58 t ii
consumption growth 46.6 7.0 24.1 22.5
investment growth 16.1 2.3 8.2 7.9
output growth 45.4 6.2 23.1 22.3
log hours 45.3 5.5 20.0 25.3
inflation 49.0 7.0 23.8 25.2
interest rate 52.1 7.1 24.9 27.2
• Estimated technology shock process:
log, technology shock at aat 1
‘recent information’
t t 11 t 2
2 t 33 t 4
4
‘earlier information’
t 55 t 6
6 t 77 t 8
8
Centered 5-quarter moving average of shocks
Signals 5-8 quarters inpast
Current shock plus most recentFour quarters’ signals
NBER peak
NBER trough
Implications for Monetary Policy• Estimated monetary policy rule induces over-
reaction to signal shock
• Problem: – positive signal induces expectation that consumption
will be high in the future
– Ramsey-efficient (‘natural’) real rate of interest jumps
– Under Taylor rule, real rate not allowed to jump, so monetary policy is expansionary
• Intuition easy to see in Clarida-Gali-Gertler model
The standard New-Keynesian Model
at at 1 t t p at log, technology
rrt rr 1 at t 1 p (natural (Ramsey) rate)
t E t t 1 xt t (Calvo pricing equation)
xt rt E t t 1 rrt E txt 1 (intertemporal equation)
rt E t t 1 xxt (policy rule)
The standard New-Keynesian Model
at at 1 t t p at log, technology
rrt rr 1 at t 1 p (natural (Ramsey) rate)
t E t t 1 xt t (Calvo pricing equation)
xt rt E t t 1 rrt E txt 1 (intertemporal equation)
rt E t t 1 xxt (policy rule)
0.95, 1.5, x 0.5, 0.82
Response to signal that technology will expand 1% in period 1
Equilibrium Ramsey
Period Period
Case Where Signal is False
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0 0 0 0
logA t 0 0 0 0 0 0 0 0
loght 0.7 0 0 0 0 0 0 0
logyt 0.7 0 0 0 0 0 0 0
Case Where Signal is True
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0
logA t 0 1 .95 .9025 0 1 .95 .9025
loght 0.7 -0.04 -0.04 -0.04 0 0 0 0
logyt 0.7 1.0 0.9 0.9 0 1 .95 .9025
0.95, 1.5, x 0.5, 0.82
Response to signal that technology will expand 1% in period 1
Equilibrium Ramsey
Period Period
Case Where Signal is False
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0 0 0 0
logA t 0 0 0 0 0 0 0 0
loght 0.7 0 0 0 0 0 0 0
logyt 0.7 0 0 0 0 0 0 0
Case Where Signal is True
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0
logA t 0 1 .95 .9025 0 1 .95 .9025
loght 0.7 -0.04 -0.04 -0.04 0 0 0 0
logyt 0.7 1.0 0.9 0.9 0 1 .95 .9025
0.95, 1.5, x 0.5, 0.82
Response to signal that technology will expand 1% in period 1
Equilibrium Ramsey
Period Period
Case Where Signal is False
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0 0 0 0
logA t 0 0 0 0 0 0 0 0
loght 0.7 0 0 0 0 0 0 0
logyt 0.7 0 0 0 0 0 0 0
Case Where Signal is True
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0
logA t 0 1 .95 .9025 0 1 .95 .9025
loght 0.7 -0.04 -0.04 -0.04 0 0 0 0
logyt 0.7 1.0 0.9 0.9 0 1 .95 .9025
0.95, 1.5, x 0.5, 0.82
Response to signal that technology will expand 1% in period 1
Equilibrium Ramsey
Period Period
Case Where Signal is False
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0 0 0 0
logA t 0 0 0 0 0 0 0 0
loght 0.7 0 0 0 0 0 0 0
logyt 0.7 0 0 0 0 0 0 0
Case Where Signal is True
0 1 2 3 0 1 2 3
4 t -1 0 0 0 0
logA t 0 1 .95 .9025 0 1 .95 .9025
loght 0.7 -0.04 -0.04 -0.04 0 0 0 0
logyt 0.7 1.0 0.9 0.9 0 1 .95 .9025
• Let’s see how a signal that turns out to be false works in the full, estimated model.
1. In the equilibrium, inflation is
below steady state2. In Ramsey, inflation
has a zero steady state
Problem: monetary policy does not raise the
interest rate enough
Price of capital (marginal cost of equity) rises in equilibrium
Sticky wages exacerbate the
problem
• The following slide corrects the hours worked response in the previous slides, which was graphed incorrectly.
Why is the Boom-Bust So Big?
• Most of boom-bust reflects suboptimality of monetary policy.
• What’s the problem?
– Monetary policy ought to respond to the natural (Ramsey) rate of interest.
– Relatively sticky wages and inflation targeting exacerbate the problem
Policy solution• Modify the Taylor rule to include:
– Natural rate of interest (probably not feasible)– Credit growth– Stock market– Wage inflation instead of price inflation.
• Explored consequences of adding credit growth and/or stock market by adding Bernanke-Gertler-Gilchrist financial frictions.
Conclusion• Estimated a model in which agents receive advance
information about technology shocks.
• Advance information seems to play an important role in business cycle dynamics
– Important in variance decompositions
– Boom-bust of late 1990s seems to correspond to a period in which there was a lot of initial optimism about technology, which later came to be seen as excessive
• Monetary policy appears to be overly expansionary in response to signal shocks
– Ramsey-efficient allocations require sharp rise in rate of interest, which `standard monetary policy does not deliver’.
– Problem is most severe when wages are sticky relative to prices.